Three Estimates, Three Different Questions
A first-in-human dose can be informed by toxicology, pharmacology, or both. These evidence sources answer different questions. A dose based on the no-observed-adverse-effect level (NOAEL) asks what animal dose produced no observed adverse effect in a particular study. A dose based on the highest non-severely toxic dose (HNSTD) asks what dose stayed below severe toxicity in a study designed to characterize that boundary. A minimum anticipated biological effect level (MABEL) asks what dose is expected to produce only a minimal level of biological activity in humans.
Because the estimates have different meanings, comparing their numerical values is useful but not sufficient. A low NOAEL-based human-equivalent dose (HED) may be influential if the toxicology species is pharmacologically relevant and the findings translate to humans. A low MABEL may be more informative when small exposures could produce substantial pharmacology, as may occur with some immune agonists. HNSTD is often considered in oncology development when severe toxicity, rather than the absence of any adverse effect, is the relevant toxicology boundary.
A NOAEL is not proof that a dose is harmless: it means that the study did not detect treatment-related adverse effects at that tested dose. Conversely, an HNSTD may include non-severe adverse effects; it is not a “no-effect” dose. MABEL is a prediction rather than an observed human threshold, so it depends on how well the evidence and human exposure predictions are supported.
Put Candidate Doses on Comparable Terms
Animal NOAEL and HNSTD values are commonly converted to HEDs before comparison with a human dose estimate. For a body-surface-area conversion, let \(K_m\) denote the species-specific conversion factor. The calculation is:
Multiplying the HED in mg/kg by a stated human body weight gives a total dose in mg. This conversion is a scaling step, not a safety factor and not evidence that animal and human responses are identical. The human dose corresponding to a MABEL is instead obtained by translating the pharmacological target into an exposure or dose using an appropriate human pharmacokinetic prediction. The form of that calculation depends on the exposure measure and dosing conditions.
To compare candidates meaningfully, make their basis explicit: state the endpoint, species or evidence source, route, units, and assumptions. Ensure that dose amounts being compared refer to compatible dosing conditions. A daily animal dose should not be compared casually with a human single dose if the dosing intervals or exposure profiles differ.
Worked Example: Comparing Three Candidate Doses for a High-Risk Biologic
Hypothetical data. A biologic has a rat NOAEL of 6 mg/kg/day. For this illustration, use \(K_{m,\mathrm{rat}}=6\), \(K_{m,\mathrm{human}}=37\), and a 70 kg adult. A dog study identifies an HNSTD of 3 mg/kg/day; use \(K_{m,\mathrm{dog}}=20\). Pharmacology and human PK evidence indicate that an unbound Cmax of 0.020 mg/L corresponds to the predefined minimal-effect target. The predicted human unbound Cmax is 0.005 mg/L per mg of the intended IV dose. Assume this relationship is linear over the dose range considered.
NOAEL-based HED. Applying the stated species conversion:
For a 70 kg adult, this corresponds to:
HNSTD-based HED. Using the dog value and the supplied factors:
The corresponding total dose is \(1.622\ \mathrm{mg/kg/day}\times70\ \mathrm{kg}=113.5\ \mathrm{mg/day}\).
MABEL-based candidate. Under the assumed linear relationship, divide the target unbound Cmax by the predicted Cmax per mg:
The three candidates are therefore approximately 68.1 mg/day from the rat NOAEL, 113.5 mg/day from the dog HNSTD, and 4.00 mg from the MABEL calculation. These are not automatically interchangeable: the first two are animal-to-human dose conversions, while the third uses a human pharmacology target and predicted human exposure. Given this hypothetical high-risk mechanism and the stated evidence, the 4.00 mg MABEL-based candidate would be the most directly aligned with limiting initial biological activity. This comparison does not apply a safety factor or, by itself, establish a final starting dose.
How the Evidence Supports a Choice
The calculation identifies candidate values; the risk assessment determines how much weight to place on them. A useful comparison starts by asking whether each estimate is based on a relevant endpoint and a credible translation to humans. For a NOAEL, examine the study design, dose spacing, duration, species relevance, and whether the adverse-effect assessment could detect the toxicity of concern. A NOAEL from a short study may not inform a risk that emerges only after prolonged exposure. A small study may also fail to detect an uncommon adverse effect.
For an HNSTD, verify how “severe” toxicity was defined and whether the study included doses that bracketed the severe-toxicity boundary. If the highest tested dose produced no severe toxicity, that dose may be an HNSTD only in the limited sense of the highest dose tested without severe toxicity; it does not identify where severe toxicity begins. HNSTD is most informative when the study and endpoint match the development context in which it is being used.
For a MABEL, check whether the target is genuinely minimal and supported by relevant pharmacology. The estimate should account for the biological system and evidence that inform the human effect, not simply select a familiar in vitro potency value without translation. Check that the exposure-to-dose mapping uses a suitable human PK prediction, the correct route, and a concentration measure compatible with the pharmacology evidence. If the prediction assumes linearity, ask whether that assumption is plausible over the range used.
A straightforward derivation illustrates the dose calculation when the target and predicted exposure use the same metric. If predicted exposure \(E\) changes linearly with dose \(D\), so that \(E=kD\) for a positive proportionality constant \(k\), then the dose corresponding to target exposure \(E_{\mathrm{target}}\) is found by solving for \(D\):
This result is valid only under the assumed proportional relationship and compatible units. If exposure is nonlinear, or if the target and model use different exposure measures, this simple calculation is not justified; the exposure-to-dose relationship must instead reflect the relevant pharmacokinetics.
Worked Example: When a Relevant NOAEL Is More Informative
Hypothetical data. A small molecule is intended for chronic dosing. In a pharmacologically relevant animal species, a well-conducted 90-day study identifies a NOAEL of 0.20 mg/kg/day. Use \(K_{m,\mathrm{animal}}=3\), \(K_{m,\mathrm{human}}=37\), and a 70 kg adult. A separate, less well-supported pharmacology translation gives a MABEL-based candidate of 8.0 mg/day. The animal findings include an organ toxicity considered relevant to human risk; the MABEL estimate depends on an assay with uncertain translation.
The NOAEL-based HED is:
For a 70 kg adult:
The NOAEL-based candidate is therefore about 1.14 mg/day, lower than the 8.0 mg/day MABEL-based candidate. Here, the NOAEL is supported by a relevant species, an adequate-duration study, and a toxicity of concern. It is reasonable for this toxicology evidence to carry substantial weight in the comparison. The MABEL should not be ignored, but its uncertain assay translation makes the point estimate less secure. This example does not establish a universal rule that the lower number always wins; it illustrates why evidence quality and relevance matter alongside the arithmetic.
Worked Example: Using HNSTD Evidence in an Oncology Context
Hypothetical data. A cytotoxic oncology drug produces treatment-related adverse findings at all tested doses, so the study does not establish a NOAEL. In a rat study, 6 mg/kg/day is the highest dose without severe toxicity; 12 mg/kg/day produces severe toxicity. For illustration, use \(K_{m,\mathrm{rat}}=6\), \(K_{m,\mathrm{human}}=37\), and a 70 kg adult. No reliable MABEL-based candidate has been established because the available pharmacology data do not support a quantitative minimal-effect target.
The 6 mg/kg/day HNSTD converts to:
The corresponding dose for a 70 kg adult is:
The HNSTD-based HED is about 68.1 mg/day. It is an informative toxicology candidate here because a NOAEL is unavailable and the study distinguishes the highest dose without severe toxicity from a dose that causes severe toxicity. It does not show that 68.1 mg/day is safe in humans: species differences remain, and this calculation does not apply a safety factor. The missing MABEL is also a limitation, not evidence that pharmacological risk is absent.
Diagnostics, Alternatives, and Common Pitfalls
A candidate becomes less dependable when the evidence does not support the assumptions needed to interpret it. Signs to investigate include a NOAEL at the highest tested dose with little dose range, an HNSTD based on a study that never approached severe toxicity, or a MABEL that changes markedly when a different assay or exposure assumption is used. Disagreement among estimates is not, by itself, a calculation error; it may reveal that toxicology and pharmacology are describing different hazards.
| Approach | Primary question | Important limitation |
|---|---|---|
| NOAEL-based HED | What animal dose had no observed adverse effect in the study? | Depends on study sensitivity, tested doses, and relevance of the species and endpoint. |
| HNSTD-based HED | What was the highest dose without severe toxicity? | May include non-severe adverse effects and may not locate the severe-toxicity boundary precisely. |
| MABEL-based dose | What dose is predicted to produce minimal biological activity in humans? | Depends on pharmacology evidence and the exposure-to-dose translation. |
A common mistake is to choose the smallest numerical estimate automatically and label it “the correct starting dose.” The lower estimate may be a sensible conservative constraint, but it may also rest on a weak or mismatched endpoint. Another mistake is to treat HNSTD as equivalent to NOAEL: a study can show non-severe adverse effects at the HNSTD. The correction is to name what each estimate means, assess its relevance and uncertainty, and state why it should or should not influence the decision.
These approaches are complementary rather than competing formulas for the same quantity. NOAEL and HNSTD are anchored in observed animal toxicology; MABEL is anchored in an anticipated human pharmacological response. A reasoned comparison can identify the estimate most relevant to the drug’s risk profile and document the evidence that remains uncertain. Converting an animal dose to an HED does not resolve those uncertainties, and this comparison is distinct from applying safety factors after HED conversion.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- A rat NOAEL is 3.7 mg/kg/day. Using \(K_{m,\mathrm{rat}}=6\) and \(K_{m,\mathrm{human}}=37\), calculate the HED in mg/kg/day.
- In a study, the highest tested dose produces no severe toxicity, but the study did not test a higher dose. What can and cannot be concluded about the HNSTD?
- A MABEL target is 0.030 mg/L, and predicted exposure is 0.010 mg/L per mg of dose. Assuming linearity, calculate the candidate dose.
- Why is a NOAEL not proof that a dose has no risk in humans?
Solutions. (1) \(3.7\times(6/37)=0.600\) mg/kg/day. (2) The highest tested dose without severe toxicity can be reported as the observed HNSTD for that study, but the severe-toxicity boundary is not established because no higher dose was tested. (3) \(0.030\ \mathrm{mg/L}\div0.010\ \mathrm{mg/L\ per\ mg}=3.00\) mg. This result relies on the stated linear relationship. (4) A NOAEL means no adverse effect was observed at that dose in the specified animal study; study sensitivity, species differences, and translation to humans limit what it establishes about human risk.